Guidance for Selecting Model Options in the National Cancer Institute Joinpoint Regression Software

Abstract

Background The National Center for Health Statistics (NCHS) often evaluates and reports on trends over time based on data from NCHS surveys and the National Vital Statistics System. The National Cancer Institute’s (NCI’s) Joinpoint software has often been used for performing trend analyses, although the software includes several options for adjusting the default menu options (for example, log transformations, modeling method and constraints, and the various model selection methods), which can lead to differing conclusions about the trend. Objective The purpose of this report is to provide guidance to users of NCHS data in the selection of modeling options when using the NCI Joinpoint regression software to analyze trends. This report complements another report, “National Center for Health Statistics Guidelines for Analysis of Trends.” Considerations are presented for selecting the modeling options, with examples illustrating the choices. The tradeoffs and consequences of choosing the various modeling options using data from NCHS data systems are discussed.